The same two price lows can show disagreement on RSI, none on the main MACD series and disagreement again on MACD−Signal. This need not be a platform error. The formulas differ, and “histogram” does not always identify the same quantity. Check the selected series, units and point-pairing rule before interpreting the pattern, then distinguish the pivot’s bar from the time it can be recognized.

What divergence actually is

Divergence compares price extremes with readings from a selected series, such as RSI, the Relative Strength Index, or MACD, the Moving Average Convergence Divergence. RSI compares smoothed positive price changes with smoothed absolute negative changes; it does not simply count rising and falling candles. Its standard scale is 0–100. A period of 14 means 14 chart periods, not necessarily 14 days.

For MACD using EMA 12 and EMA 26, the main series is their difference. The signal line smooths that series. MT5 documents SMA 9 for the signal; TradingView allows EMA or SMA choices and defines its histogram as MACD−Signal. MT5’s built-in MACD draws the main series as bars. Thus “12, 26, 9” and the shape of a bar plot do not establish identical calculations across providers. Check series names and values in the data window.

The four types of divergence worth knowing

Regular bullish geometry pairs a lower price low with a higher oscillator low; regular bearish pairs a higher price high with a lower oscillator high. Hidden geometry reverses those relations and is commonly studied as a continuation hypothesis within a specified trend context. Names and interpretations vary: Cardwell discussions also use positive and negative reversals. The separate guide to hidden versus regular divergence develops these comparisons; neither category certifies the next move.

Four selected-point geometries
Regular bullishPrice: lower low. Selected oscillator: higher low. A reversal hypothesis, not a promised rise.
Regular bearishPrice: higher high. Selected oscillator: lower high. A reversal hypothesis, not a promised fall.
Hidden bullishPrice: higher low. Selected oscillator: lower low. A continuation hypothesis in an upward context.
Hidden bearishPrice: lower high. Selected oscillator: higher high. A continuation hypothesis in a downward context.

Hidden divergence is conventionally interpreted as a possible continuation setup, while regular divergence suggests a possible reversal. Neither category has universally higher expectancy. Test a precise definition and execution rule on a stated sample, including costs.

The same price points, different indicator readings

Use a hypothetical EUR/USD quote in USD per EUR, with one pip = 0.0001. Read each indicator exactly at the recognized price pivots at bars 10 and 20. The supplied EMA, signal and RSI values are assumed states calculated from earlier history; that history is not supplied, so this table does not reconstruct their initialization. It also does not establish that these bars are independent oscillator minima.

Hypothetical readings at two price lows
QuantityA: bar 10B: bar 20B versus A
Price low, USD per EUR1.10001.098020 pips lower
RSI, scale points304010 points higher
EMA 12, USD per EUR1.10301.1007MACD input
EMA 26, USD per EUR1.10401.1020MACD input
Main MACD = EMA 12 − EMA 26−0.0010−0.00130.0003 lower
Signal line−0.0006−0.0011Specified smoothing states
Histogram = MACD − Signal−0.0004−0.00020.0002 higher

Price falls and RSI rises: the exact-bar comparison has bullish disagreement. Main MACD falls and does not. MACD−Signal rises because its absolute gap to the signal narrows; both values remain negative. “Higher” does not mean “above zero.” These are three transformations of the same price history, not automatically three independent observations.

Timeframes and price units change different things

Timeframe changes the observations and trading horizon. It does not by itself prove greater reliability. Define comparable divergence rules for each interval and evaluate their outcomes, trading costs and sample size before ranking them.

Do not compare MACD amplitudes with RSI scale points as if they shared units. In our technical scaling test, multiply every input price and initial average by 100, keeping periods unchanged. The EMA recurrence scales its output by 100; main MACD, its averaged signal and their difference scale by 100 too. Values −0.0010 and −0.0013 become −0.10 and −0.13 with the same ordering.

Both RSI components U and D scale together, leaving U/D unchanged: 0.0003/0.0007 and 0.03/0.07 still give RSI 30. The same selected-point classification is preserved, but a MACD difference threshold of 0.0002 must become 0.02; a 10-point RSI threshold is unchanged. This is a data transformation, not a profitability result. Reciprocal quotation 1/price is not a constant multiplier and does not follow this simple scaling argument.

A pivot is recognized after its price bar

  1. Define the detector. In our own model, a low must be strictly below the lows of two bars on either side; ties fail. A marker’s historical position does not show when it became known.
  2. Supply the neighbouring data. Lows at bars 18–22 are 1.1005, 1.0990, 1.0980, 1.0988 and 1.0995. Bar 20 qualifies only after bar 22 closes; A was recognized earlier after bar 12.
  3. Keep the timing honest. Do not book an entry at bar 20 using confirmation from bar 22. Record recognition, order and fill separately. If bar 21’s low were 1.0975, bar 20 would fail this detector.

How to manage a divergence trade

Choose a stop relative to the setup's invalidation point and measured volatility. ATR multipliers and pip distances are candidates for testing, not universal settings. An ordinary stop can slip, and ATR alone cannot establish the probability that it will be hit.

If a strategy uses a level from drawing support and resistance, specify its construction before the outcome, together with entry, exit, size and costs. A stop-to-entry adjustment is a rule to test rather than a compulsory next step, and an ordinary stop at entry does not guarantee a cost-free result. Divergence alone does not set a safe volume or stop distance.

Keep separate records for comparison and execution
InputInstrument, provider, source, timeframe, formula, periods and smoothing types.
Point pairingExact price-pivot bars or independently chosen oscillator extrema; do not switch after seeing the result.
RecognitionLatest required bar and whether the input bar is closed.
ObservationFixed evaluation horizon and every qualifying case, including those without reversal.
Trade testOrder and fill assumptions, sizing, exits and costs, if a trading rule is evaluated.
Unused evaluationTest a fixed version without retuning; extra filters do not automatically create an edge.

Flat prices, different indicator responses

This models two flat closes after earlier movement. No pivots or signals are sought; only smoothing mechanisms are compared.

Scroll sideways to see the full diagram

Flat prices, different indicator responsesInitial EMA(3)=105 and EMA(7)=110. Two new closes of 100 give EMAs 102.5/107.5, then 101.25/105.625. Main MACD is −5, −5, −4.375. RSI(3), starting U=1 and D=2 with zero changes, preserves U/D=1/2 and a reading of 33.33.After price stops, averages still changeTwo successive closes: 100, 100P = 100Main MACD(3,7):−5−5−4.375RSI(3): U and D decay proportionally33.3333.33Initial stateAfter 2 bars

Hypothetical model: initial states are specified, not reconstructed from two candles. EMA uses α = 2/(n + 1); RSI uses Wilder smoothing with n = 3. Each series has its own scale; amplitudes are not compared between panels. Formulas: EMA in MT5 and RSI in TradingView.

MACD can change without a new price move on the same bar.

Smoothing can change after price stops

Divergence describes a disagreement between price extremes and an oscillator; it does not guarantee reversal.

The diagram assumes initial EMA(3) = 105 and EMA(7) = 110, then two closes of 100. Using α = 2/(n + 1), the averages become 102.5/107.5 and then 101.25/105.625. Main MACD is −5, −5, then −4.375 even though the new closes are unchanged. RSI(3) starts with U = 1 and D = 2; two zero changes multiply both components by 2/3 at each step, preserving U/D = 1/2 and RSI ≈ 33.33. These specified states illustrate memory, not pivots or a measured trading signal. There is no universal two- or three-bar RSI lead over MACD.

MACD compares moving averages and RSI compares smoothed gains and losses, so their responses can differ. Simultaneous divergence is another rule to test; it does not automatically strengthen a signal, and the two indicators are not independent evidence.

What to do tomorrow

  1. Change only the signal. Keep table B’s main MACD at −0.0013 but set its signal to −0.0008. The difference becomes −0.0005, lower than A’s −0.0004: bullish disagreement on MACD−Signal disappears, while RSI 30 → 40 is unchanged.
  2. Check units. Multiply all prices and starting average states by 100. Recalculate MACD and thresholds without multiplying RSI scale points.
  3. Check availability. Record why bar 20 requires bar 22 and how a lower bar-21 low invalidates it. Drawing the marker retrospectively does not make it available earlier.
  4. Design a fixed evaluation. Declare the selected series, pairing, horizon and complete execution rules. Include every qualifying case and costs, and reserve unused data. No universal timeframe or compulsory number of patterns is required.